Trends in chemical composition of global and
regional population-weighted fine particulate matter over the recent 25 years
Chi Li ([email protected]), PhD Candidate
Randall V. Martin, Aaron van Donkelaar, Brian L. Boys, Melanie S. Hammer, Junwei Xu,
Eloise A. Marais, Adam Reff, Madeleine Strum, David A. Ridley, Monica Crippa, Qiang Zhang
8th International GEOS-Chem Meeting
Various materials to study PM2.5 composition trends
In situ measurements
• No global coverage
Model simulation
• Needed for global study of composition trends.
• Bias in population-weighted mean (PWM) PM2.5 and estimated mortality due to dilution
into coarse grid cells alone (Li et al., 2016)
Satellite-based estimates of PM2.5
• Combine satellite AOD with modeled PM2.5/AOD
• No speciation information, lack of insight into
sources and processes(Hand et al., 2012)
(Boys et al., 2014)
Downscaling GEOS-Chem to 0.1˚ × 0.1˚ based on satellite-derived PM2.5 provides
better representation of population-weighted mean (PWM) PM2.5 exposure
Before downscaling After downscaling
v11, MERRA2, 1989-2013, time-
varying inventories, aqueous
isoprene SOA (Marais et al., 2016)
Scale each composition by
PM2.508−12(satellite)
PM2.508−12(model) 1989-2013
mean PM2.5 (μg/m3)
PW
M P
M2.5
(μg
m-3
)
satellite-based PM2.5 estimates
Species mass
fraction and
relative trend is
consistent with the
original simulation
Long-term decreases in PWM PM2.5 over North America driven by sulfate and OA
Simds: downscaled simulation (collocated with obs)(PM2.5 at 35% RH and species are dry)
**: p < 0.05
*Negligible PWM trends from dust and sea salt
*Similar conclusions for short-term trends (2002-2013) based on more stations
Long-term decreases in PWM PM2.5 over North America driven by sulfate and OA
Simds: downscaled simulation (collocated with obs)(PM2.5 at 35% RH and species are dry)
**: p < 0.05
*Negligible PWM trends from dust and sea salt
*Similar conclusions for short-term trends (2002-2013) based on more stations
Long-term decreases in PWM PM2.5 over North America driven by sulfate and OA
Simds: downscaled simulation (collocated with obs)(PM2.5 at 35% RH and species are dry)
**: p < 0.05
*Negligible PWM trends from dust and sea salt
*Similar conclusions for short-term trends (2002-2013) based on more stations
Promising consistency with satellite-derived PM2.5 trends over 1998-2013
Significantly (p < 0.05) increasing global
PWM PM2.5, with overlapping 95%
confidence interval (CI, in square brackets).
Regional PWM trends over 21 Global Burden
of Diseases Study (GBD) regions:
Trendsat: satellite-based PM2.5 estimates
Simds: downscaled simulation
PM2.5 at 35% RH
**: p < 0.05
(Boys et al., 2014;
Van Donkelaar et al., 2015)
Promising consistency with satellite-derived PM2.5 trends over 1998-2013
Significantly (p < 0.05) increasing global
PWM PM2.5, with overlapping 95%
confidence interval (CI, in square brackets).
Regional PWM trends over 21 Global Burden
of Diseases Study (GBD) regions:
Trendsat: satellite-based PM2.5 estimates
Simds: downscaled simulation
PM2.5 at 35% RH
**: p < 0.05
(Boys et al., 2014;
Van Donkelaar et al., 2015)
+ ++
++
_ __
+
O
OO
O
+ Both significant increase (p < 0.1)
- Both significant decrease (p < 0.1)
O Both insignificant (p ≥ 0.1)
O
O
Promising consistency with satellite-derived PM2.5 trends over 1998-2013
Significantly (p < 0.05) increasing global
PWM PM2.5, with overlapping 95%
confidence interval (CI, in square brackets).
Regional PWM trends over 21 Global Burden
of Diseases Study (GBD) regions:
Trendsat: satellite-based PM2.5 estimates
Simds: downscaled simulation
PM2.5 at 35% RH
**: p < 0.05
(Boys et al., 2014;
Van Donkelaar et al., 2015)
20 regions with overlapping 95% CI except
Oceania
+ ++
++
_ __
+
O
OO
O
+ Both significant increase (p < 0.1)
- Both significant decrease (p < 0.1)
O Both insignificant (p ≥ 0.1)
O
O
Significant trends over populous regions driven by sulfate, nitrate, ammonium, OA
*: p < 0.1
**: p < 0.05
Pie charts: species PWM concentration (PM2.5 in middle) over 89-13
Bar plots: species PWM trends (filled, p < 0.1; blank, p ≥ 0.1) over 89-13 (left) and 98-13 (right)
Thick blue: satellite-based PM2.5 estimates
(PM2.5 and species at 35% RH)
Significant trends over populous regions driven by sulfate, nitrate, ammonium, OA
*: p < 0.1
**: p < 0.05
Pie charts: species PWM concentration (PM2.5 in middle) over 89-13
Bar plots: species PWM trends (filled, p < 0.1; blank, p ≥ 0.1) over 89-13 (left) and 98-13 (right)
Thick blue: satellite-based PM2.5 estimates
(PM2.5 and species at 35% RH)
Significant trends over populous regions driven by sulfate, nitrate, ammonium, OA
*: p < 0.1
**: p < 0.05
Pie charts: species PWM concentration (PM2.5 in middle) over 89-13
Bar plots: species PWM trends (filled, p < 0.1; blank, p ≥ 0.1) over 89-13 (left) and 98-13 (right)
Thick blue: satellite-based PM2.5 estimates
(PM2.5 and species at 35% RH)
Global PWM trends: sharp difference with area-weighted mean (AWM) trends
• Global increase (μg m-3yr-1) in PWM PM2.5 (0.28) driven by OA (0.10), nitrate (0.05),
sulfate (0.04), ammonium (0.03), dust (0.03) and BC (0.02) over 1989-2013.
• Weak AWM PM2.5 trends vs. significantly increasing PWM trends, not reflective of
population exposure.
**: p < 0.05 PWM AWM
(Li et al., ES&T, in prep)
Pie charts: species PWM concentration (PM2.5 in middle) over 89-13
Bar plots: species PWM trends (filled, p < 0.1; blank, p ≥ 0.1) over 89-13 (left) and 98-13 (right)
Thick blue: satellite-based PM2.5 estimates
Satellite-based estimates of PM2.5 help constrain simulated PWM PM2.5.
Significant PWM PM2.5 trends driven by OA and secondary inorganic
aerosols globally, and over densely populated regions, consistent with in situ
trends over North America.
PWM trend is stronger and more insightful on PM2.5 exposure than AWM
trend.
Thank you!
Summary
Acknowledgements
Extra slides for Q/A
RegionInventory
(coverage)Used species Annual scale factor Reference
World
EDGAR v4.3.1
(1970-2010)
CO, NOx, SO2,
NH3, OC, BCN/A Crippa et al. (2016)
RETRO
(2000)VOCs
from EDGAR v4.3.1,
1970-2010Schultz (2007)
USEPA NEI
(2011)
CO, NOx, SO2,
NH3, OC, BC,
VOCs
NEI historical emission,
1990-2014
US Environmental
Protection Agency
CanadaCAC
(2002-2008)
CO, NOx, SO2,
NH3, OC, BCAPEI, 1990-2014 Environment Canada
MexicoBRAVO
(1999)CO, NOx, SO2
from EDGAR v4.3.1,
1970-2010Kuhns et al. (2005)
EuropeEMEP
(1990-2012)CO, NOx, SO2, NH3 N/A
Centre on Emission
Inventories and
Projections
AsiaMEIC
(2008-2012)
CO, NOx, SO2,
NH3, OC, BC,
VOCs
from EDGAR v4.3.1 +
SO2, OC, BC from Lu et al.
(2011), 1970-2012
Li et al. (2017)
Time-varying emissions
• Open fire emissions: RETRO (1981-1996, scaled by GFED4 based on 1997-2000 ratio) and GFED4 (1997-2014)
• Anthropogenic Emissions
Stronger decreases in European sulfate before 2000, especially over
Central Europe• Excluding the Po Valley site yields:
Obs: -0.15 μg m-3 (-1.4%) yr-1
Simds: -0.23 μg m-3yr-1 (-1.6%) yr-1
Global OA trends driven dominantly by POA and OPOA
POAOPOATSOAASOAISOA
Seasonal trends in OA over North AmericaW
inte
rS
um
me
r
Seasonal driver of OA trends (POA/OPOA in winter and Isoprene SOA in
summer) over North AmericaW
inte
rS
um
me
rPOA OPOA TSOA ASOA ISOA
Trends in PWM PM2.5 over barren areas driven by dust/open fire
**: p < 0.05
Pie charts: species PWM concentration (PM2.5 in middle) over 89-13
Bar plots: species PWM trends (filled, p < 0.1; blank, p ≥ 0.1) over 89-13 (left) and 98-13 (right)
Thick blue: satellite-based PM2.5 estimates
(PM2.5 and species at 35% RH)
Recent increase in wood burning and OC emission over the eastern US
causes underestimation of OA decreases over the US
Winter
Summer
(μg m-3yr-1)
Solid: total emissions
Dashed: anthropogenic emissions
Global trends in population-weighted mean (PWM) PM2.5 driven by secondary
inorganic aerosols (SIA) and OA (1989-2013)
**: p < 0.05
Regional PM2.5 trends
driven mainly by:
OA and sulfate over India
Nitrate and OA over China
Sulfate over the US and
Europe
OA over Amazon and
African rainforest
Dust over North Africa
and Arabic Peninsula
Difference in scales for
PM2.5 and composition
Contrasts in PWM vs. AWM PM2.5 over different regions
AWMPWM
Only nitrate shows significant increase over East Asia, 2005-2012
Consistent with Geng et al. (2017)
Calculation of OC/EC emissions over North America
𝐎𝐂 =
𝐬𝐞𝐜𝐭𝐨𝐫𝐬
𝐏𝐌𝟐.𝟓 ∙ 𝐎𝐂%
𝐄𝐂 =
𝐬𝐞𝐜𝐭𝐨𝐫𝐬
𝐏𝐌𝟐.𝟓 ∙ 𝐄𝐂%
US total emission (Gg)
NEI 11 This work
OC 520 586
BC 259 306
• Emission factors for different sectors from SPECIATE
database (Reff et al., 2009).
• Canada BC emission of 42 Gg comparing to 43 Gg
for 2014 in a recent EC inventory.
